基于普鲁克分析的润滑油红外光谱指纹区相似度评价方法

The Method of Similarity Determination for Lubricating Oil Infrared Fingerprint by Procrustes Analysis

基于普鲁克分析的润滑油红外光谱指纹区相似度评价方法,能够有效地降低由于实验条件的差异导致的同一润滑油样品红外光谱在峰幅值大小、峰位置偏移等方面的差异。采集了42个不同来源、不同种类润滑油样品的红外光谱,将其作为标准样品;对其中22个样品在不同实验条件(不同实验人员和涂片方式)下重复采集其红外光谱,作为未知待识别样品。识别结果证明,该相似度评价方法能够较好地实现润滑油的鉴别分析,相似度值的大小能够真实地反映润滑油样品红外光谱指纹区之间的相似性;通过控制相似度值的范围(大于0.9),能在一定程度上实现润滑油的种类识别。

The new similarity evaluation algorithm between lubricating oils based on Procrustes analysis （PA） could reduce the difference in peak amplitude and position shift of the infrared fingerprint area resulted from experiment conditions.Infrared spectra of 42 lubricating oil samples from different sources and of types were collected as standard samples.The infrared spectra of 22 samples from the 42 samples were collected at the different experimental conditions and then regarded as ‘unknown＇ samples to be identified.The experimental results proved that lubricating oil identification could be well accomplished based on the new algorithm,and the type identification of lubricating oil could be done well if similarity coefficient was greater than 0.9.